Supported Models¶
Zwischen supports reviewer models from four providers. Set the model key in your settings file to the model ID.
Model list¶
| Provider | Model ID | Notes |
|---|---|---|
| Anthropic | claude-opus-4-7 |
Most capable Anthropic model |
claude-sonnet-4-6 |
Balanced speed and capability | |
claude-haiku-4-5 |
Fastest Anthropic model | |
gemini-3.1-pro-preview |
Latest Gemini preview | |
gemini-2.5-pro |
Strong general-purpose model | |
gemini-2.5-flash |
Fast and cost-effective | |
gemini-3.1-flash-lite |
Lightest Gemini model | |
| OpenAI | gpt-5.5 |
Most capable OpenAI model |
gpt-5.4 |
Balanced performance | |
gpt-5-mini |
Fast and cost-effective | |
| DeepSeek | deepseek-v4-pro |
Most capable DeepSeek model |
deepseek-v4-flash |
Fast DeepSeek model |
Authentication¶
Each provider requires its own API key, set via api_key in settings or the provider's environment variable:
| Provider | Environment variable |
|---|---|
| Anthropic | ANTHROPIC_API_KEY |
GEMINI_API_KEY |
|
| OpenAI | OPENAI_API_KEY |
| DeepSeek | DEEPSEEK_API_KEY |
OpenAI also supports OAuth authentication (auth_mode: oauth). See Models and Auth.
Model tuning¶
Different models behave differently as reviewers. Zwischen Labs maintains provider-specific and model-specific prompt adjustments that account for these behavioral differences. These are applied automatically when model_tuning: true (the default).
Set model_tuning: false to disable adjustments and use the base reviewer prompt.